Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Steve Jobs: Xserve Axed Over Poor Sales
- Thread starter MacRumors
- Start date
- Sort by reaction score
Actually, the turtlenecked one's "big picture" is focussed on "thin" above all.
The Steve-inspired project to make a ½U XServe recently foundered, so the "petulant one" cancelled the whole XServe program since he couldn't announce "the world's thinnest datacenter server".
</sarcasm>
Yes, two Mini macs per 6 square foot of datacenter floor space would make any datacenter manager happy.
Not.
Did you forget part of your post (if nothing else, the "sarcasm" tag)?
I've said it before, but... Do HP worry about the profit they make on Printers? Of course not... their profit is on toner, as Apple's is on iMac etc etc.
10,000 servers / quarter doesn't sound that bad... Plus the boost that must be coming from corporate iPad adoption etc.
It's not like they're broke now is it?
No, they don't, that is a ridiculously silly thing to say.
Have you ever tried binding Macs against a Linux Open Directory server? Good luck with that.
Your mini goes down and you're on vacation. You've got two spare paper weights on the shelf.
Failover isn't easy to setup.. and to do it right you've got another, expensive, load balancer in front.
You obviously don't work in enterprise.
You use iSCSI! Wait, only one gigabit port... well.. er... you use USB Ethernet dongles for front side access and the gigabit for the storage. That's the ticket.
Virtual adapters? Wireless N? I'd recommend neither but that's how.
Mac mini "server" is a kids toy when compared to Xserve. Xserve is a server that lot of the post production (moving image) shops have used. We have couple of them (with couple of Xserve Raids) and we have no complaints. We will never buy some workstation hardware like "MacPro server" or toys like Mac mini server. Honestly, couple of thousand is nothing when it comes to robust servers. Downtime or loss of data would cost us far far more then that.
So your complaint is iSCSI vs. Fiber Channel? I think actually getting the Minis onto a SAN is accomplishment enough to not discuss performance, don't you?
Not a single one.
Last edited by a moderator:
Well they probably aren't because they are good at deciding what to sell and what not to sell. Whenever Apple or any other company discontinues a product, it always causes issues and people deal with it. This is not the first time such thing happens. Remember Shake? And that was even a good selling product.
That figure was from several years ago. It's probably less than that now, especially with the scaled back spending on the backend during this recession by large enterprises. Even if Apple is selling 10,000 per quarter, that's not even close to having 1% market share.
http://www.zdnet.com/blog/microsoft/idc-windows-server-still-rules-the-server-roost/6424
I'm not a techie but I follow the industry trends in non-Mac sites and the Xserve and OS X Server seem to get laughed at by the "serious" Unix, Linux, and Windows IT guys.
Mac mini is freaking frisbee not a server. Its a toy! You honestly think we would rely on using Mac mini "server" when dealing with projects worth >$100 000?
Yup and EDUs and Design Firms don't need 1000 Xserves.. a MacPro form factor works fine if you only need 5-10 machines.
iPhone is managed in the cloud, and considering the strategy of all of Apple's competitors is to focus on cloud services...I expect the same from Apple.
No, but I have tried to get Windows clients to use a Linux Samba file server.
The operative word is "tried". After fighting the half-assed CIFS implementation in Linux for more than a week I realized that the "path of least resistance" was to use the native NFS (AKA the "Nightmare File System) support in Windows to mount NFS exports from the Linux boxes.
Later, I re-imaged the file servers to run Windows Server 2008 - and left all the problems behind.
And if I ever again hear someone say that "Linux can be the PDC for your domain" I will probably erupt into violence on the spot.
What would the point of virtualization be in this situation? There's nothing that you can't colocate on a single iteration of OSX, is there?
I doubt that figure has dropped with the increase in the Apple market share these past few years. More likely it has INCREASED. And it's still 100% of the market share in pure 100% mac deployments. It has to be.
Apple sells lots of iphones... home users buy them.
Apple sells lots of iPads... home users buy them.
Home users buy more apple macs... MacBook, iMac... blah.
Home users go in to work and rave about their iPhone / iPad / iMac / Macbook .... corporates buy them...
Corporate sysadmins see how many of their users are demanding iOS . OSX devices... Need and OSX server to manage them....
Whoops... best we can offer is some pikey Mac Pros 'Under a desk someplace'.
Cue hysterics at Apples' amatuerish '6U per server' offerings...
The end.
MacPro's lack the redundancy that is required from the server. Honestly, if you rely on Xsan and Final Cut Server you won't be looking at having a MacPro when it comes to server hardware. Next stop, Apple kills Xsan and Final Cut Server...
It's funny how people claim that this will be "bad business" for Apple in the end, that those XServes disappearing will result in more macs disappearing from the workplace etc etc.
If this was a devastating decision for Apple, do you really think Apple wouldn't know about that?
Get real.
Apple knows exactly how much business they are going to lose from this, you don't. And if Apple thinks it's still worth to axe the entire product, then eventually they are going to profit from it, not lose money.
If this was a devastating decision for Apple, do you really think Apple wouldn't know about that?
Get real.
Apple knows exactly how much business they are going to lose from this, you don't. And if Apple thinks it's still worth to axe the entire product, then eventually they are going to profit from it, not lose money.
That Youtube page had a related link to CNBC's graphics SAN using Apple XServe which quickly explains why neither Mac Pros or Mini macs are useful replacements for the XServe
For the datacenter neophytes - watch that clip to understand how absurd Apple's "migration path" really is.
If you think that USB is a "CPU hog", don't even think about how much CPU will be consumed running iSCSI on a mini or Mac Pro !!
Serious iSCSI (by "serious" I mean just about anything but a home system) needs server-class NICs with dedicated iSCSI offload engines to avoid killing the CPU. (A typical NIC has a couple of 64-bit CPUs to handle the network and iSCSI protocol stacks).
It's also "best practice" to have a completely separate network for the iSCSI NICs (isolated from the house and world networks) with separate jumbo-frame supporting wire-speed Gigabit switches.
On a mini - haha.
Dude, you really shouldn't be taking that discussion too seriously - I was only playing devil's advocate as to how you could connect a SAN. And to continue, Airport Wireless-N for the seperate network.
Moreover, have you ever tried to integrate your macs to an OD / AD golden triangle when your server platform is Linux?
You can replicate management on Linux. OD is, at it's core, LDAP. You can write your own MCX files and push them to your clients.
This is all an enormous amount of work that I can accomplish with a few clicks in OS X server.
Now given benefits generally cost 50% on top of salary.. Is it good or bad business move for a company to buy a $5K XServe instead of buying a $4k Linux box so you can have your expensive support staff spend endless hours managing a hand-rolled configuration? Even if you already have an LDAP server.. $5k buys a little over 2 weeks of my time with benefits. I work in EDU.. If I was a mid-Sr. Unix admin in the private sector, I'd probably be a more expensive.
BTW, there is no production-ready AD support for Linux other than through purchasing 3rd party apps (more cost).. and they make me nervous. Apple proved that it'd difficult to integrate into AD if you're not Microsoft.
Bottom line..
A server is a server.
A workstation is a workstation.
A desktop is a desktop.
and a mini is a toy.
A Mac Pro doesn't even qualify as a workstation, and certainly not a tower server.
- It's not rack mountable because the handles make it too wide for a 19" rack.
- It doesn't have tool-less power supplies and motherboards like any other major workstation vendor. Swapping MBs and PSs is downright difficult.
- It doesn't have redundant power supplies as an option like any enterprise server should. Even a decent SMB tower server should have a hot swap, redundant option in the product line.
- HP will ship me parts over night with a base workstation warranty. Dell will sign me up for Warranty Direct for free if I piggy back enrollment on a large order and they'll actually pay me to replace my own parts (over-nighted). Apple will insist on sending out a 3rd party repair company if I've got Extended Apple Care. They're supposed to be there next day, no 4 hour service available. (they've generally been good when we need them, but not always next-day)
There's a lot of posts here from people who have obviously never provided enterprise support. If you worked for me and you bought a desktop or a toy to run mission critical services, you'd deserve to get fired.
We pay good money for enterprise features because downtime is money. It can be A LOT OF MONEY, even in Edu when you've got distance learning in a VERY expensive master's program and the people in Singapore need to see the class stream. Even in the markets where people buy Apple servers (not the Board of Trade or Amazon..) a nights downtime can be worth tens of thousands of dollars.. easily.
As to where Apple's going..
I'm often wrong in guessing what's going on in Steve's head but the logical move is to allow licensing for VMWare ESX. OS X Server licensing already extends to VMWare Fusion (on Mac hardware of course).
Apple could easily announce a return to higher OS X Server license pricing for VMWare users. I'd be happy as a clam if I could run OS X Server in my VMWare ESX environment and sink the money that I spend on XServes into my virtualization infrastructure. I might even run more OS X Server.
Server hardware is so damn fast these days that most of my cycles sit idle anyway.. even in a ESX environment running 10 servers (mail, DB, web, filesharing...)
Unfortunately, I can see Jobs balking at this.
Just couldn't get into the market and gave up before wasting $.MacPro's lack the redundancy that is required from the server. Honestly, if you rely on Xsan and Final Cut Server you won't be looking at having a MacPro when it comes to server hardware. Next stop, Apple kills Xsan and Final Cut Server...
Apple's been pushing Towers for meta-data controllers for a long time because they're redundant ( you really should have at least two), they don't take much horsepower, and customers balk and spending 5K for a MDC.
Towers do make some sense here because the system is designed from the ground up to be redundant.
Too bad they don't make an earless pro that would sit on it's side in a rack.
separate different services onto different (virtual) servers
Assuming that you mean "host" rather than "colocate" (since "colocation" has a distinct meaning referring to off-site network servers) - industry "best practices" suggest running one service per server.
In practical terms, this means running one service per VM - hosted on the appropriate number of physical servers depending on the load each service presents.
Instead of putting all your eggs in one basket, run your DNS server in one VM, your DHCP server in a second, your web server in a third, your file server in a 4th, your....
If you have to upgrade or you have a problem with your file server - you can debug or reboot file services without any effect on DNS, DHCP or web services.
So yes, as you suggest, you could run 10 services on one host. Many say, however, that running the 10 services in 10 VMs is a better idea (even if the 10 VMs are on that same host).
Assuming that you mean "host" rather than "colocate" (since "colocation" has a distinct meaning referring to off-site network servers) - industry "best practices" suggest running one service per server.
In practical terms, this means running one service per VM - hosted on the appropriate number of physical servers depending on the load each service presents.
Instead of putting all your eggs in one basket, run your DNS server in one VM, your DHCP server in a second, your web server in a third, your file server in a 4th, your....
If you have to upgrade or you have a problem with your file server - you can debug or reboot file services without any effect on DNS, DHCP or web services.
So yes, as you suggest, you could run 10 services on one host. Many say, however, that running the 10 services in 10 VMs is a better idea (even if the 10 VMs are on that same host).